Any Backgammon players here? In a recent thread, a Doper asked about Go and Backgammon. Backgammon is a good game to take up: the games are often very exciting. The GNU Backgammon software probably plays better than any human and can be downloaded for free. It will point out your mistakes. Just now I asked it to analyze my recent match against a human at PlayOK.com. In the position below it told me that my move (6-2 as X) was “Very Bad,” even though it was the 2nd-best move.
Can you guess the best move?
The score (after 5 games) is: O 6, X 5 (match to 7 points), Crawford game
Date: 5/7/2018
Place: PlayOK
GNU Backgammon Position ID: nHsiAwRrNpEBMA
Match ID : 8AnrAGAAKAAE
+13-14-15-16-17-18------19-20-21-22-23-24-+ O:
| X O O | | O O O X | 6 points
| X | | O O O X |
| | | O O O |
| | | O |
| | | |
| |BAR| | 7 point match (Cube: 1)
| | | |
| | | |
| | | |
| O | | X X X X | Rolled 62
| O X X | | X X O X X X | 5 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X:
Pip counts: O 114, X 125
X to play 62
Do some of you still select the “Sultan-theme Break Code Tags” option at lower-right? If so, the display inside Spoiler might work.
The score (after 5 games) is: O 6, X 5 (match to 7 points), Crawford game
Date: 5/7/2018
Place: PlayOK
GNU Backgammon Position ID: nHsiAwRrNpEBMA
Match ID : 8AnrAGAAKAAE
+13-14-15-16-17-18------19-20-21-22-23-24-+ O:
| X O O | | O O O X | 6 points
| X | | O O O X |
| | | O O O |
| | | O |
| | | |
| |BAR| | 7 point match (Cube: 1)
| | | |
| | | |
| | | |
| O | | X X X X | Rolled 62
| O X X | | X X O X X X | 5 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X:
Pip counts: O 114, X 125
X to play 62
So you rolled a 6-2? This is an interesting one. 8-6, 11-5 seems like the most obvious move. But it leaves you with two annoying anchors at 24 and 13 with few options for flexible play.
I’d probably end up doing something like 24-18 and 13-11. Yes, I do split up two anchors to form one anchor, but that seems the most flexible position to me. I also considered 13-7, 13-11, but taking my checkers off the 13 point leaves little flexibility for my checkers on 24 to get over, and I’m basically conceding his checkers on 12 to just make a run for the home board.
Anyhow, that’s my top-of-the-head thought. Seeing how well GNU backgammon plays against me, though, I’m sure this is completely wrong.
BTW, the actual match situation, 6-5 Crawford, meant that I could gammon him but he couldn’t really gammon me, and the cube was out of play for the rest of the match. GNU Backgammon knows this when it made its analysis.
However, I reran the position with match score 0-0 and, although the rankings of the moves changed, the same unlikely-looking move was still ranked as best by a very large margin. (The values of the moves in rank order were -.16, -.44, -.52, -.57, -.94, -.96, -1, -1, -1, … — make any move other than one of the best six and you will immediatley fold a double.)
I’ll be curious to hear the GNU analysis and what the unlikely looking move is. Ah, I just saw one more move that makes some sort of sense: 8-2 and 13-11. That still strikes me as a bit conservative. I’m going to stick to my seemingly wreckless 24-18, 13-11. That still looks like the best move to me. Although 8-2, 13-11 does leave him in a bit of a pickle if he happens to get hit, with five of the six inner board points covered. But that one’s not really unlikely looking, so it can’t be the one being suggested by the computer.
Reckless? How about 8-2, 6-4? If you can survive the opponent’s next roll, you might be in line for a gammon. I can never get my mind around it, but I’m always amazed at the computer’s willingness to lose a piece on its inner table.
Ah, that’s an interesting one that I didn’t even think of. That doesn’t feel like a good move to me, but it does look intriguing for the reason you stated. But leaving two exposed points seems like quite the gamble. That would qualify as an unlikely looking move, though.
Here are the 12 best moves, according to GNU Backgammon, given the 6-5 Crawford score:
(1) 8/2 6/4* +.32
(4) 13/11 8/2 -.05
(2) 11/9 8/2 -.10
(3) 11/5 8/6 -.11
(-) 24/18 8/6 -.11
(-) 24/18 6/4* -.14
(-) 24/16 -.16
(-) 13/7 6/4* -.16
(-) 11/5 6/4* -.18
(6) 13/11 13/7 -.24
(-) 24/18 11/9 -.24
(5) 11/3 -.26
The number in parentheses is the rank order when the problem is presented as the 1st game of a 7-point match.
Sometimes, when beaten by a good player, I run GNU Backgammon to see what my mistakes were. In this match, I led 5-0 but ended up losing 5-7. :mad: GNU showed only a few of my moves as Very Bad (much fewer than opponent’s). I try to learn … but the correct move in the problem position is one I’d *never *have gotten right!
Well, there you go! I was completely wrong! Good job to Ass for a Hat for spotting the best move! And 8-6/11-5 (the one I called the “obvious” move) was surprisingly much better than I thought it would be, ranking #4 and #3 in the two scenarios. And, looking at it again, 8-2, 13-11 should have also struck me as an obvious move. :smack: But I definitely would have missed the winning move in any case.
In my defense, I will say that 24-18/13-11 is probably being strongly influenced by that being my usual opening for a 6-2 roll, so I thought in this game state it would still be somewhat favorable (even though it leaves a blot on 13 which doesn’t exist in the opening, and it leaves that blot on 8. I don’t much care about the blot on 18), but it appears not even to show up in the top ten!
I have to admit, I’d probably wouldn’t have the stones to make this move if I were actually playing. I always end up being waaaay too optimistic about the position I’m in to play high risk moves like this one.
Me too. That is a real tough one to find ‘over the board’ - I can’t believe many human players, even experts, would make that move, even if they considered it. The problem is that even if oppo misses the double-shot, there are no rolls that give you the perfect board needed to make a gammon likely. The best roll is probably something with a 2, to cover one blot with the other, and that leaves the 6-point open. Or possibly you roll two 7s in a row with oppo fanning twice (perfect scenario). It all seems a bit unlikely, but GNU does the roll-outs and I don’t, so I’m prepared to believe it. The most surprising thing to me is that it is still the best move at 0-0; at 5-6 down, clearly gammons are massively more valuable, which is not the case at 0-0. I’d be interested to see if it was the same ranking at DMP, actually.
